Deliver: Insights Feedback Loop - #57
Merged
Merged
Conversation
Author: architect Waves: 3 Tasks: 4 Out-of-scope deps: no
Plan: .agents/plans/insights-feedback-loop.md (Status + approved event written by: node harness/cli.js approve)
Wave 1, Task 1.1 Validated: AC#1, AC#2 — node --test green; append-only diff verified Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…rt A) Wave 2, Task 2.1 Validated: AC#3 — documented Step 4c one-liner run against the real repo (noWaveData zeros path) and a synthetic events.jsonl fixture (non-zero outcome/retry/fail-reason/veto/spend counts) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…tion + env knob (Part B) Wave 2, Task 2.2 Validated: AC#4, AC#5 — documented Step 3b computation run against the real findings.jsonl (default threshold 5 lists exactly testing/code-clarity/security/error-handling/correctness; RAD_FINDINGS_THRESHOLD=100 lists none; unset/0/NaN/negative all parse to 5); .env.example block matches the existing commented doc-block style Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Wave 3, Task 3.1 Validated: AC#6 — ran the documented Step 4b computation against the real .agents/findings.jsonl: default threshold 5 yields testing, code-clarity, security, error-handling, correctness; RAD_FINDINGS_THRESHOLD=100 yields nothing (line omitted) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Delivers the approved insights-feedback-loop plan — the feedback-loop piece of the Agent Reliability Stack review. Pure read-side: no new event types, no log writes, gate fold untouched.
What landed (3 waves, 4 tasks, all 7 ACs)
Wave 1 — read helpers (Part A foundation)
harness/events.jsbelowtotalUsage, same tolerance contract:outcomeCounts,failReasonCounts,retryCounts,hookVetoCounts. Writer/fold code byte-identical (append-only diff verified). Unit tests cover counting from spine-shaped synthetic histories plus the zeros contract on empty/approved-only logs.Wave 2 — /rad-insights sections
.agents/state/*/events.jsonlcross-feature via the events.js helpers — success rate by the frozen 7-outcome vocabulary, retries, failure-reason distribution, hook vetoes, token spend per wave — with an explicit "no wave data yet" zeros rendering (validated against the real repo, which has no wave events yet, and against synthetic wave data).RAD_FINDINGS_THRESHOLD, default 5) each emit a ready-to-paste CLAUDE.md convention line or described lint rule, framed "suggestion — apply via PR; never auto-applied". Today's data triggers: testing (14), code-clarity (12), security (9), error-handling (7), correctness (5). Env knob documented in.env.example.Wave 3 — /wrap touchpoint (droppable tail)
Verification
check-scope.sh: 5/5 changed files in declared scope ·check-tests.sh: pass · harness suite: 0 failurescheck-approval-integrity.sh: full PASS locally (ancestry, fingerprint, gate, authenticity) — this PR is the first delivered end-to-end under the new CI.agents/logs/insights-feedback-loop-2026-07-03.md🤖 Generated with Claude Code